Persian Gulf Interior Runs the Highest-Confidence Approach Encounters in the Rule 17 Dataset. Its Deviation Rate Is a Third of Rotterdam's.

The Setup

Every close-quarters vessel encounter that crosses into give-way territory gets tagged by context: open sea, anchorage, or approach — the water between open sea and the berth, where a vessel is inbound or outbound and traffic funnels through a defined corridor. In the last 90 days, the approach-context table alone holds 930,317 pairwise encounters, of which 45.6% show a Rule 17 deviation — the stand-on vessel maneuvering before the give-way vessel forces a handoff.

Broken out by port, most approach zones cluster in a predictable band: Rotterdam's deviation rate is 57.7%, Antwerp's is 62.9%, Singapore's is 46.8%, Houston's is 35.4%. One zone doesn't fit the band. Persian Gulf Interior — the bulk-terminal approach corridor covering 22,838 encounters over the same 90 days — runs a 13.8% deviation rate, roughly a third of Rotterdam's and less than half of Houston's.

The Chain

The obvious read is a data-quality problem: sparse AIS coverage produces fewer detected close-quarters events, which produces an artificially low deviation count. The numbers say the opposite. Persian Gulf Interior's average encounter confidence is 0.778 — the highest of any approach zone in the corpus, more than double Rotterdam's 0.337, Singapore's 0.411, Antwerp's 0.379, and Houston's 0.335. Only 5.0% of Persian Gulf Interior encounters fall below the 0.3 confidence floor; at every comparison port, 40–49% do. This is the best-observed approach corridor in the dataset, and it's also the one with the fewest Rule 17 deviations.

The trigger mix offers a partial explanation. Of Persian Gulf Interior's 22,838 approach encounters, 58.1% resolve via "extremis" — a last-moment handoff after the give-way vessel has failed to act — versus 41.9% via ordinary give-way inaction. At Rotterdam, extremis accounts for 78.1% of handoffs. A corridor with proportionally fewer extremis resolutions is a corridor where more encounters get sorted out before they become emergencies, which is consistent with — though not proof of — a genuinely lower deviation rate rather than a detection gap.

The week-by-week pattern adds a second layer. The earliest full week in the window (July 27) ran 34.3% deviation on 4,057 encounters — close to the global approach average. Every week since has run 8–12%, with one outlier week (August 24, 31.1% on just 716 encounters) breaking the pattern. The rate didn't drift down gradually; it stepped down and mostly stayed there.

The Implication

A signal this clean and this anomalous is exactly the kind of thing an anchor-vessel-conduct model should not average away. If Persian Gulf Interior's low deviation rate holds, it means give-way behavior in this corridor is materially different from every other bulk or container approach zone Overwatch tracks — worth understanding on its own terms, not folding into a global Rule 17 baseline where it would just look like noise pulling the average down.

What to Watch

Three things determine whether this is a real behavioral pattern or an artifact of a young zone: whether the deviation rate holds as weekly sample sizes stabilize above 1,000, whether the extremis-to-inaction ratio continues to run below Rotterdam's, and whether average confidence stays above 0.7 as more vessel classes enter the corridor. A reversion toward the 35–63% band seen elsewhere would suggest the current numbers are a small-sample effect rather than a durable regional pattern.

Limitations

Persian Gulf Interior as a distinct approach zone was created on March 25, 2026, and there is no pairwise-encounter data tagged to it before the 90-day window analyzed here — there is no prior-period baseline to compare against, only the within-window week-over-week series. Weekly volume swings from 132 to 8,177 encounters, and the most recent week is a partial week (132 encounters, likely undercounted). Encounter confidence measures AIS position-report density and consistency, not navigational skill or intent — a high-confidence, low-deviation reading says the underlying vessel-track data is trustworthy, not that a specific crew handled a specific encounter well. This is a directional finding on two months of stable pattern, not a multi-year baseline.

Data as of 2026-09-08. Source: pairwise_encounter, port_zones, ports (Axiom Overwatch), 90-day trailing window.